Premium
Uncertainty assessment in watershed‐scale water quality modeling and management: 1. Framework and application of generalized likelihood uncertainty estimation (GLUE) approach
Author(s) -
Zheng Yi,
Keller Arturo A.
Publication year - 2007
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2006wr005345
Subject(s) - glue , watershed , uncertainty analysis , sensitivity analysis , computer science , scale (ratio) , probabilistic logic , hydrological modelling , watershed management , uncertainty quantification , environmental science , data mining , engineering , machine learning , artificial intelligence , simulation , geography , mechanical engineering , cartography , climatology , geology
Watershed‐scale water quality models involve substantial uncertainty in model output because of sparse water quality observations and other sources of uncertainty. Assessing the uncertainty is very important for those who use the models to support management decision making. Systematic uncertainty analysis for these models has rarely been done and remains a major challenge. This study aimed (1) to develop a framework to characterize all important sources of uncertainty and their interactions in management‐oriented watershed modeling, (2) to apply the generalized likelihood uncertainty estimation (GLUE) approach for quantifying simulation uncertainty for complex watershed models, and (3) to investigate the influence of subjective choices (especially the likelihood measure) in a GLUE analysis, as well as the availability of observational data, on the outcome of the uncertainty analysis. A two‐stage framework was first established as the basis for uncertainty assessment and probabilistic decision‐making. A watershed model (watershed analysis risk management framework (WARMF)) was implemented using data from the Santa Clara River Watershed in southern California. A typical catchment was constructed on which a series of experiments was conducted. The results show that GLUE can be implemented with affordable computational cost, yielding insights into the model behavior. However, in complex watershed water quality modeling, the uncertainty results highly depend on the subjective choices made by the modeler as well as the availability of observational data. The importance of considering management concerns in the uncertainty estimation was also demonstrated. Overall, this study establishes guidance for uncertainty assessment in management‐oriented watershed modeling. The study results have suggested future efforts we could make in a GLUE‐based uncertainty analysis, which has led to the development of a new method, as will be introduced in a companion paper. Eventually, the study should assist in the development of a new generation of watershed water quality models.